Search

As more and more online search queries come from voice, automatic speech recognition becomes a key component to deliver relevant search results. Errors introduced by automatic speech recognition (ASR) lead to irrelevant search results returned to the user, thus causing user dissatisfaction. In this paper, we introduce an approach, Mondegreen, to correct voice queries in text space without depending on audio signals, which may not always be available due to system constraints or privacy or bandwidth (for example, some ASR systems run on-device) considerations. We focus on voice queries transcribed via several proprietary commercial ASR systems. These queries come from users making internet, or online service search queries. We first present an analysis showing how different the language distribution coming from user voice queries is from that in traditional text corpora used to train off-the-shelf ASR systems. We then demonstrate that Mondegreen can achieve significant improvements in increased user interaction by correcting user voice queries in one of the largest search systems in Google. Finally, we see Mondegreen as complementing existing highly-optimized production ASR systems, which may not be frequently retrained and thus lag behind due to vocabulary drifts., Comment: Accepted in KDD 2021

Many recent advances in neural information retrieval models, which predict top-K items given a query, learn directly from a large training set of (query, item) pairs. However, they are often insufficient when there are many previously unseen (query, item) combinations, often referred to as the cold start problem. Furthermore, the search system can be biased towards items that are frequently shown to a query previously, also known as the 'rich get richer' (a.k.a. feedback loop) problem. In light of these problems, we observed that most online content platforms have both a search and a recommender system that, while having heterogeneous input spaces, can be connected through their common output item space and a shared semantic representation. In this paper, we propose a new Zero-Shot Heterogeneous Transfer Learning framework that transfers learned knowledge from the recommender system component to improve the search component of a content platform. First, it learns representations of items and their natural-language features by predicting (item, item) correlation graphs derived from the recommender system as an auxiliary task. Then, the learned representations are transferred to solve the target search retrieval task, performing query-to-item prediction without having seen any (query, item) pairs in training. We conduct online and offline experiments on one of the world's largest search and recommender systems from Google, and present the results and lessons learned. We demonstrate that the proposed approach can achieve high performance on offline search retrieval tasks, and more importantly, achieved significant improvements on relevance and user interactions over the highly-optimized production system in online experiments., Comment: Accepted at CIKM 2020

Reinforcement learning methods for recommender systems optimize recommendations for long-term user engagement. However, since users are often presented with slates of multiple items---which may have interacting effects on user choice---methods are required to deal with the combinatorics of the RL action space. We develop SlateQ, a decomposition of value-based temporal-difference and Q-learning that renders RL tractable with slates. Under mild assumptions on user choice behavior, we show that the long-term value (LTV) of a slate can be decomposed into a tractable function of its component item-wise LTVs. We demonstrate our methods in simulation, and validate the scalability and effectiveness of decomposed TD-learning on YouTube.

Generalized linear models with nonlinear feature transformations are widely used for large-scale regression and classification problems with sparse inputs. Memorization of feature interactions through a wide set of cross-product feature transformations are effective and interpretable, while generalization requires more feature engineering effort. With less feature engineering, deep neural networks can generalize better to unseen feature combinations through low-dimensional dense embeddings learned for the sparse features. However, deep neural networks with embeddings can over-generalize and recommend less relevant items when the user-item interactions are sparse and high-rank. In this paper, we present Wide & Deep learning---jointly trained wide linear models and deep neural networks---to combine the benefits of memorization and generalization for recommender systems. We productionized and evaluated the system on Google Play, a commercial mobile app store with over one billion active users and over one million apps. Online experiment results show that Wide & Deep significantly increased app acquisitions compared with wide-only and deep-only models. We have also open-sourced our implementation in TensorFlow.

Bigtable is a distributed storage system for managing structured data that is designed to scale to a very large size: petabytes of data across thousands of commodity servers. Many projects at Google store data in Bigtable, including web indexing, Google Earth, and Google Finance. These applications place very different demands on Bigtable, both in terms of data size (from URLs to web pages to satellite imagery) and latency requirements (from backend bulk processing to real-time data serving). Despite these varied demands, Bigtable has successfully provided a flexible, high-performance solution for all of these Google products. In this article, we describe the simple data model provided by Bigtable, which gives clients dynamic control over data layout and format, and we describe the design and implementation of Bigtable.

We study the consensus problem, which requires multiple processes with different input values to agree on one of these values, in the context of asynchronous shared memory systems. Prior research focussed either on t-resilient solutions of this problem (which must be correct even if up to t processes crash) or on wait-free solutions (which must be correct despite the crash of any number of processes). In this paper, we show that these two forms of solvability are closely related. Specifically, for all $n > t \ge 2$ and all sets ${\mathcal{S}}$ of shared object types (that include simple read/write registers), there is a t-resilient solution to n-process consensus using objects of types in ${\mathcal{S}}$ if and only if there is a wait-free solution to (t + 1)-process consensus using objects of types in ${\mathcal{S}}$. Our proof of this equivalence uses another result derived in this paper, which is of independent interest. Roughly speaking, this result states that a wait-free solution to (n - 1)-process consensus is never necessary in designing a wait-free solution to n-process consensus, regardless of the types of objects available. More precisely, for all $n \ge 2$ and all sets ${\mathcal{S}}$ of shared object types (that include simple read/write registers), if there is a wait-free solution to n-process consensus that uses a wait-free solution to (n - 1)-process consensus and objects of types in ${\mathcal{S}}$, then there is a wait-free solution to n-process consensus that uses only objects of types in ${\mathcal{S}}$.

We describe efficient algorithms for projecting a vector onto the l1-ball. We present two methods for projection. The first performs exact projection in O(n) expected time, where n is the dimension of the space. The second works on vectors k of whose elements are perturbed outside the l1-ball, projecting in O(k log(n)) time. This setting is especially useful for online learning in sparse feature spaces such as text categorization applications. We demonstrate the merits and effectiveness of our algorithms in numerous batch and online learning tasks. We show that variants of stochastic gradient projection methods augmented with our efficient projection procedures outperform interior point methods, which are considered state-of-the-art optimization techniques. We also show that in online settings gradient updates with l1 projections outperform the exponentiated gradient algorithm while obtaining models with high degrees of sparsity.

The publish/subscribe (or pub/sub) paradigm is an increasingly popular model for interconnecting applications in a distributed environment. Many existing pub/sub systems are based on pre-defined subjects, and hence are able to exploit multicast technologies to provide scalability and availability. An emerging alternative to subject-based systems, known as content-based systems, allow information consumers to request events based on the content of published events. This model is considerably more flexible than subject-based pub/sub. However, it was previously not known how to efficiently multicast published events to interested content-based subscribers within a large and geographically distributed network of broker (or router) machines. We develop and evaluate a novel and efficient distributed algorithm for this purpose, called -link matching". Link matching performs just enough computation at each node to determine the subset of links to which an event should be forwarded. We show via simulations that: link matching yields higher throughput than flooding when subscriptions are selective; and the overall CPU utilization of link matching is comparable to that of centralized matching.